Using the deep Spitzer Infrared Array Camera ( IRAC ) observations of the Great Observatories Origins Deep Survey ( GOODS ) , we study the stellar masses and star formation histories of galaxies at z \approx 6 .
The IRAC instrument provides the best opportunity to estimate the stellar masses of galaxies at these redshifts because it samples their rest-frame optical fluxes , which are less prone to dust extinction and are more sensitive to the light from longer-lived stars .
Our study is based on the i _ { 775 } -band dropout sample selected from the GOODS southern and northern fields ( \sim 330 arcmin ^ { 2 } in total ) , several of which already have spectroscopic confirmations .
In total , we derive stellar masses for 53 i _ { 775 } -band dropouts that have robust IRAC detections .
These galaxies have typical stellar masses of \sim 10 ^ { 10 } M _ { \odot } and typical ages of a couple of hundred million years , consistent with earlier results based on a smaller sample of z \approx 6 galaxies in the Hubble Ultra Deep Field .
We also study 79 i _ { 775 } -band dropouts that are invisible in the IRAC data and find that they are typically less massive by a factor of ten .
These galaxies are much bluer than those detected by the IRAC , indicating that their luminosities are dominated by stellar populations with ages \lesssim 40 million years .
We discuss various sources of uncertainty in the mass estimates , and find that our results are rather robust .
The existence of galaxies as massive as 10 ^ { 10 } M _ { \odot } at z \approx 6 can be explained by at least one set of N-body simulations of the hierarchical paradigm .
Based on our mass estimates , we derive a lower limit to the global stellar mass density at z \approx 6 .
Considering the range of systematic uncertainties in the derived stellar masses , this lower limit is 1.1 to 6.7 \times 10 ^ { 6 } M _ { \odot } Mpc ^ { -3 } ( co-moving ) , which is 0.2 to 1.1 % of the present-day value .
The prospect of detecting the progenitors of the most massive galaxies at yet higher redshifts is explored : a deep , wide-field near-IR survey using our current technology could possibly result in positive detections at z > 7 .
We also investigate the implication of our results for reionzation , and find that the progenitors of the galaxies comparable to those in our sample , even in the most optimized ( probably unrealistic ) scenario , can not sustain the reionization for a period longer than \sim 2 million years .
Thus most of the photons required for reionization must have been provided by other sources , such as the progenitors of the dwarf galaxies that are far below our current detection capability .